tool-evaluation-and-adoption
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external, untrusted data such as GitHub repositories, package registry metadata, and documentation.
- Ingestion points: references/reconnaissance.md instructs the agent to clone repositories and read files like READMEs, documentation, and architecture records.
- Boundary markers: While the skill defines clear 'evaluate-only' vs 'install-authorized' scopes, it lacks specific instructions for the agent to use delimiters or ignore instructions found within the analyzed external content.
- Capability inventory: The skill possesses capabilities for cloning repositories, running package managers (npm, pip, cargo, etc.), creating git worktrees, and executing installation commands.
- Sanitization: The skill mitigates risks by requiring the agent to 'read install scripts before running them' and 'verify checksum-verified release assets,' which are defensive best practices.
- [COMMAND_EXECUTION]: The skill instructs the agent to use standard development tools for auditing and verification.
- Evidence: Mentions using git worktrees to measure bundle impact and various package managers (npm, pip, poetry, etc.) to perform dependency audits and local installations.
Audit Metadata